2022-05-09

Data set overview

Title: “Peripheral Blood Mitochondrial DNA Copy Number Is Associated with Prostate Cancer Risk and Tumor Burden”
Authors: Zhou W. et. al. (2014)
Purpose: Predict cancer from biomarkers, mainly mtDNA

Loading

  • Dimensions: 392, 13

  • Control and cancer case groups are proportional

Cleaning

  • Check for duplicates

  • Filter for PCR success (pcr_success)

  • New dimensions: 387, 13

Augmenting

  • BMI- and DFI-classifier

  • New columns based on TNM-notation

  • Add ‘group’ as strings

  • New dimensions: 387, 18

Flowchart for project flow

Boxplot with continuous variables, any outliers?

<<<<<<< HEAD

Boxplot with discrete variables, any outliers?

======= <<<<<<< HEAD

Boxplot with discrete variables, any outliers?

=======

Boxplot with discrete variables, any outliers?

>>>>>>> 61340d8cffed6dbb557e481f18015a4b16ee4f64 >>>>>>> a7bcc6a58d0b4e55b85e40426a66e61e13eae94e

Re-creating plot from the article

A better biomarker for prostate cancer?

<<<<<<< HEAD

======= <<<<<<< HEAD

=======

>>>>>>> 61340d8cffed6dbb557e481f18015a4b16ee4f64 >>>>>>> a7bcc6a58d0b4e55b85e40426a66e61e13eae94e

Some exploratory data analysis

Logistic regression, excl. PSA

Significant p-values:
Maybe the distribution of DFI-classes are skewed?

Logistic regression, incl. PSA

Significant p-values:

Principal component analysis (PCA)

Interesting finding during exploratory data analysis

Conclusion

  • We can support the conclusion of the article, mtDNA is a biomarker for prostate cancer and cancer severity (e.g, it is reproducible)
  • PSA levels are better for cancer prediction
  • Both of the above could be supported by Logistic Regression and PCA